Summary

The current forecast date is 05 January, showing data until 08 January.

The main forecasting model is an unweighted ensemble of three individual models:

  1. unweighted time series ensemble model (autoregressive ARIMA, ETS and naive models)
  2. regression + 7-day-lagged cases (cases mapped from UTLA to Trust using Trust-UTLA mapping in covid19.nhs.data)
  3. convolution from cases to admissions (scaled; cases mapped to Trusts as above)

The time series ensemble and regression models are trained on the last 6 weeks of data from 24 November; the case-convolution model is trained on the last 9 weeks of data (from 03 November), of which 6 are used in the likelihood. Forecasts are made for the following 14 days.

For details of the fitted models, see our preprint “Comparative assessment of methods for short-term forecasts of COVID-19 admissions in England at the local level”.

Current forecasts

Trusts, all regions

Summary, table

Current forecast, ribbon

Forecast increase, scatter

Forecast increase, maps

Trusts, by NHS region

Note: for clarity, we only show forecasts for Trusts which have admitted five or more patients in the last 7 days.

East of England

London

Midlands

North East and Yorkshire

North West

South East

South West

National and regional

We derive forecasts of national and regional hospital admissions from the Trust-level admissions. For each of the individual models, the forecast samples are summed to get national/regional forecasts, then summarised into quantile forecasts, and then the ensemble forecast created. These are compared below to forecasts made using EpiNow2.

National (England)

Regional

Focus Trusts

We compare last weeks forecasts (from 29 December 2021) to the current observed data.

Current week

Last week

Past forecasts

Trusts, all regions

National

We derive forecasts of national hospital admissions from the Trust-level admissions. For each of the individual models, the forecast samples are summed to get national forecasts, then summarised into quantile forecasts. The ensemble forecast is created from the national quantile forecasts.

Regional

We derive forecasts of regional hospital admissions from the Trust-level admissions. For each of the individual models, the forecast samples are summed to get regional forecasts, then summarised into quantile forecasts. The ensemble forecast is created from the regional quantile forecasts.